AI Scout Daily
News

AMD Is Buying Fei-Fei Li's World Labs for $8.2 Billion in Stock. What It Means for Marble and 3D World Models.

The all-stock deal makes Li AMD's chief scientist and pulls one of the leading world-model labs toward AMD's chip roadmap. Neither company has said what changes for Marble's users.

Priya Nair·September 29, 2026·4 min read

TL;DR

AMD has agreed to acquire World Labs, the spatial-intelligence lab co-founded by Fei-Fei Li, in an all-stock deal valued at about $8.2 billion, expected to close by the end of 2026 subject to regulatory approval. World Labs builds models that generate, reconstruct, and simulate interactive 3D environments from text, images, and video; its product Marble creates persistent 3D worlds. Li becomes AMD's executive vice president and chief scientist, reporting to CEO Lisa Su, and co-founders Justin Johnson and Ben Mildenhall will keep leading the team. Neither company has said anything about Marble's pricing, API, or roadmap after the deal.

AMD's largest AI move of the year isn't a chip. It is a research lab whose models generate 3D worlds, and whose founder helped start the modern computer-vision era with ImageNet. The announcements focus on AMD's hardware roadmap. For people who use World Labs' tools, the more immediate question is what happens to them, and neither company has answered it.

The deal

  • Price: about $8.2 billion, paid entirely in AMD stock.
  • Timing: announced September 28; expected to close by the end of 2026, subject to regulatory approvals.
  • People: Fei-Fei Li becomes AMD executive vice president and chief scientist, reporting to Lisa Su. Justin Johnson and Ben Mildenhall will work with Li to keep leading the World Labs team.
  • Stated purpose: AMD says World Labs' model expertise will show it how AI workloads are changing and shape its hardware, software, and systems roadmap for an "open ecosystem".

How it got here

WhenWhat happened
2024World Labs launched; The Verge reports it was valued at $1 billion within months
2025Marble, its first commercial product, launched; World Labs and AMD began a training and inference partnership on AMD GPUs
Early 2026Li appeared at AMD's CES presentation
July 2026World Labs discussed its SceniX acquisition and robotics simulation
Sep 28, 2026Definitive agreement for AMD to acquire World Labs
From AMD, World Labs, The Verge, and TechCrunch.

What World Labs actually makes

World Labs describes itself as building frontier models that perceive, generate, reason about, and interact with virtual and physical worlds. Marble creates spatially coherent, persistent 3D worlds from images, video, text, and 3D layouts, and is pitched both for entertainment and for simulated environments to train robots. Its investor list includes both AMD and Nvidia. TechCrunch notes that Nvidia already offers open-weight world models through Cosmos, while AMD has so far offered only text and video models publicly.

What hasn't been said

  • Whether Marble, its API, and current pricing continue unchanged after the deal closes.
  • Whether future World Labs models will be released as open weights. World Labs says the combined team is committed to "widely accessible open models", and Li wrote on her Substack, as quoted by The Verge, that the work will be done "openly with the community", but no specific model or license was named.
  • Whether World Labs' models will be optimized for AMD hardware only, or stay hardware-neutral.

What this means if you use AI to make images, video, or 3D

Nothing changes before the deal closes, and existing Marble projects aren't affected by the announcement itself. If Marble is part of a production workflow, keep your exports local and watch for terms or pricing updates after closing. The bigger signal is the category: with AMD buying World Labs and Nvidia backing Cosmos, both chipmakers now treat 3D world generation as core infrastructure rather than a niche creative tool, which should mean more investment in these models, and a stronger pull toward each company's hardware.

PN
Priya Nair

AI Industry Reporter

Priya covers model releases, industry announcements, and the gap between what labs claim and what independent evaluators actually find. She reads the primary source - the paper, the system card, the benchmark org's own statement - before writing a word.